Nonresponse Bias and Random Digit Dialing Part 2

In the last article, we introduced the problem of nonresponse bias. Today, we’ll look at the way that nonresponse bias affects random digit dialing specifically, or RDD.

Remember that nonresponse bias is often seen in:

  • Poor response rates
  • Eligibility Criteria
  • Dropout

RDD affects many of these and more in the following ways:

  • Missed Calls – RDD only dials a number once. If the person is not around and misses the call, it may be for a reason that affects your data. For example, if you’re trying to talk to the average family, and the number you call is someone that has to work at night, then you’re not getting the average family. Those that miss calls often have something in common.
  • Hang Ups – Of course, those that hang up also have something in common – namely, how little they want to answer your survey. Whether hang ups also introduce a bias is unclear, but they’re pretty common with RDD, and could increase in some systematic way. For example, it’s not unlikely that those that have their numbers unlisted or on the do not call list will hang up.
  • No Phone – Those without a phone or those with fewer phone numbers are also less likely to be reached. Simply by probability, those with multiple phone numbers will have a better shot at getting called than those with one or none.
  • Family Status – Multiple families within a single household are also excluded from the data, but they are relevant for things like voting, or purchasing habits, etc. It’s not necessarily very common, but it could easily play a role when you use random digit dialing methods.

Many of these fall under the heading of “eligibility.” Each of these represents systematic problems with getting random responses from your sample, and while most people see response rate in terms of whether or not they answer the survey, issues like not being home, being without a phone, sharing a phone, and more, all affect the eligibility of the respondent naturally, and all of them may affect the quality of your data.

You don’t need to do random digit dialing to see similar issues pop up in the rest of your research. There are issues with eligibility, especially (but not exclusively) for scientific surveys, that is introduced in your research methodologies.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>